The structure of domestic firm
production networks

EconCR 23

Alonso Alfaro-Ureña

Central Bank of Costa Rica and University of Costa Rica

Mariany Fuentes

University of Michigan

José Ignacio González

LSE

Isabela Manelici

LSE, CEP and CEPR

José P. Vásquez

LSE, CEP and CEPR

December 15, 2023

Motivation

  • COVID-19 pandemic showed that we are not prepared to face global supply chain disruptions.
  • Climate crisis and natural disasters are also a threat to local supply chains.
  • Better understanding on the transmission of shocks through the production network.
  • How can we improve the resilience of the production network? First, we must understand its structure.

The structure of domestic firm production networks

Costa Rica (2008-2019)

  • Describe the distribution of the firm size of Costa Rican domestic formal firms by industry and year:
    • Number of customers.
    • Number of suppliers.
    • Network sales.
    • Network purchases.
    • Total sales.

The structure of domestic firm production networks

Costa Rica (2008-2019)

  • Characterize firm and customer-supplier relationships patterns with stylized facts:
    • Firm-to-firm trade (9 facts).
    • The geography of firm-to-firm trade (3 facts).
    • Firm-to-firm trade within and across borders (3 facts).
  • Compare these stylized facts to those reported in other countries.

Main results

Firm-to-firm trade I

  1. The distribution of firm size within an industry shows substantial dispersion.
  2. Domestic production networks are sparse.
  3. Firm size is positively correlated to the number of connections to other firms and their average transaction.
  4. Firm size is positively correlated to the number of municipalities and number of industries with which a firm trades.
  5. Both the extensive and intensive margins significantly explain the variation of networks sales and purchases. As we aggregate from the firm- to the industry-level, the intensive margin becomes more important.

Main results

Firm-to-firm trade II

  1. Network sales (purchases) is positively correlated with the number of customers (suppliers).
  2. On average, well-connected firms tend to trade with poorly-connected firms.
  3. Top customers (suppliers) account for a high percentage of a firm’s network sales (purchases).
  4. The persistence of firm-to-firm linkages increases with firm size.

Main results

The geography of firm-to-firm trade

  1. Most firm-to-firm linkages are formed between nearby firms.
  2. Trade flows between municipalities increase in the geographical proximity between the municipalities.
  3. The number of connections decays with distance and travel time faster than the average transaction.

Main results

Firm-to-firm trade within and across borders

  1. Importer (exporter) firms are larger than non-importer (non-exporter) firms.
  2. Most firms do not import or export directly. Most firms import indirectly, and very few export indirectly.
  3. Direct and total export and import shares increase with firm size.

Contributions

  • Characterize several patterns present in the structure of the production networks.
    • Estimate every fact in a less-developed country.
    • Apply a consistent methodology, and sample selection.
    • Domestic firm production network.
  • Account for the complex geography and poor road infrastructure of Costa Rica.
    • New distance and travel time measures between municipalities using OpenStreetMap contributors (2023).

Outline

  1. Data:
    1. BCCR datasets.
    2. Sample selection.
    3. Route distance and travel time between municipalities.
  2. Stylized facts:
    1. Firm-to-firm trade.
    2. The geography of firm-to-firm trade.
    3. Firm-to-firm trade within and across borders.
  3. Conclusion.

Data

Data

BCCR datasets (2008-2019)

  1. VAT transaction data (Ministry of Finance): Annual transactions between formal firms.
  2. Firm balance sheets (Ministry of Finance): Total sales, 4-digit industry, and district-level location.
  3. Employment and wagebill (CCSS).
  4. Firm-level international trade data (Customs): Annual imports and exports by firm.
  5. Shared ownership data (BCCR).

Sample selection

  • At least one permanent worker during the year.
  • Exclude firms in the “Diplomatic Activities” industry.

Data

OpenStreetMap contributors (2023): Talamanca, Limón to Coto Brus, Puntarenas

  • Fastest route: 8 1/3 hours vs. Naïve point-to-point: 1 1/2 hours.

Data

OpenStreetMap contributors (2023)

  • \(\text{Distance}_{od}\) and \(\text{Travel time}_{od}\): Shortest path between municipalities miniziming the travel time.
  • \(\text{Distance}_{oo} = \frac{2}{3} \sqrt{\text{Area} \cdot \pi}\) (Ahlfeldt & Wendland, 2016).
  • \(\text{Travel time}_{oo} = \frac{\text{Distance}_{oo}}{\text{Speed}_{oo}}\).
  • \(\text{Speed}_{oo}\): Weighted average of lanes’ speed by its length.

Stylized facts:
Domestic firm-to-firm trade

The distribution of firm size within an industry shows substantial dispersion

The distribution of firm size within an industry shows substantial dispersion

The distribution of firm size within an industry shows substantial dispersion

The distribution of firm size within an industry shows substantial dispersion

The distribution of firm size within an industry shows substantial dispersion

Production networks are sparse

32,000-48,000 firms in a given year. ~96,000 unique firms.

On average, 1 in 5,500 of the possible connections are realized.

Firm size is positively correlated to the number of connections to other firms.

Firm size is positively correlated to the average transaction.

Well-connected firms tend to trade with poorly-connected firms.

Suppliers

Well-connected firms tend to trade with poorly-connected firms.

Customers

The persistence of firm-to-firm linkages increases with firm size.

Survival

The persistence of firm-to-firm linkages increases with firm size.

Creation

Stylized facts:
The geography of firm-to-firm trade

Flows decay with distance and travel time.

\[\log(\text{Total flows})_{ijt} = \beta \log(\tau)_{ij} + \alpha_{it} + \gamma_{jt} + \varepsilon_{ijt}\]

The number of connections decays with distance and travel time…

\[\log(N)_{ijt} = \beta \log(\tau)_{ij} + \alpha_{it} + \gamma_{jt} + \varepsilon_{ijt}\]

… faster than the average transaction.

\[\log(\text{Avg trans})_{ijt} = \beta \log(\tau)_{ij} + \alpha_{it} + \gamma_{jt} + \varepsilon_{ijt}\]

Stylized facts:
Firm-to-firm trade within and across borders

Direct and total import shares (Dhyne et al., 2021).

\[s^{\text{Total}}_{Fj} = s_{Fj} + \sum_{i}s_{ij}s^{\text{Total}}_{Fi}\]

  • \(s_{Fj} = \frac{\text{Imports}_{j}}{\text{Imports}_{j} + \text{Wagebill}_{j} + \sum_{i} \text{Transactions}_{ij}}\)
  • \(s_{ij} = \frac{\text{Transactions}_{ij}}{\text{Imports}_{j} + \text{Wagebill}_{j} + \sum_{i} \text{Transactions}_{ij}}\)
  • Solve for \(s^{\text{Total}}_{Fj}\).

Most firms do not import directly, but do so indirectly.

Direct and total export shares (Dhyne et al., 2021).

\[r^{\text{Total}}_{jF} = r_{jF} + \sum_{i}r_{ji}r^{\text{Total}}_{iF}\]

  • \(r_{jF} = \frac{\text{Exports}_{j}}{\text{Exports}_{j} + \text{Sales to households}_{j} + \sum_{i} \text{Transactions}_{ji}}\)
  • \(r_{ji} = \frac{\text{Transactions}_{ji}}{\text{Exports}_{j} + \text{Sales to households}_{j} + \sum_{i} \text{Transactions}_{ji}}\).

Most firms do not export at all.

Conclusion

Conclusion

  • We characterize several patterns present in the structure of the production networks.
    • The patterns are not random at all.
    • The extensive margin is very important in Costa Rica.
  • Measurement matters:
    • Microdata vs input-output tables.
    • Complex geography of Costa Rica.
    • Firm exposure to international trade.

Thank you!
j.i.gonzalez-rojas@lse.ac.uk

References

Acemoglu, D., & Azar, P. D. (2020). Endogenous Production Networks. Econometrica, 88, 33–82.
Ahlfeldt, G. M., & Wendland, N. (2016). The Spatial Decay in Commuting Probabilities: Employment Potential vs. Commuting Gravity. Economics Letters, 143, 125–129.
Arkolakis, C., Huneeus, F., & Miyauchi, Y. (2023). Spatial Production Networks (NBER Working Paper Series No. 30954). National Bureau of Economic Research.
Bernard, A. B., Dhyne, E., Magerman, G., Manova, K., & Moxnes, A. (2022). The Origins of Firm Heterogeneity: A Production Network Approach. Journal of Political Economy, 130, 1765–1804.
Bernard, A. B., Jensen, J. B., Redding, S. J., & Schott, P. K. (2009). The Margins of US Trade. American Economic Review, 99, 487–493.
Bernard, A. B., Moxnes, A., & Saito, Y. U. (2019). Production Networks, Geography, and Firm Performance. Journal of Political Economy, 127, 639–688.
Blaum, J., Lelarge, C., & Peters, M. (2018). The Gains from Input Trade with Heterogeneous Importers. American Economic Journal: Macroeconomics, 10, 77–127.
Chaney, T. (2014). The Network Structure of International Trade. American Economic Review, 104(11), 3600–3634.
Dhyne, E., Kikkawa, A. K., Komatsu, T., Mogstad, M., & Tintelnot, F. (2022). Foreign Demand Shocks to Production Networks: Firm Responses and Worker Impacts [Working Paper].
Dhyne, E., Kikkawa, A. K., Mogstad, M., & Tintelnot, F. (2021). Trade and Domestic Production Networks. Review of Economic Studies, 88, 643–668.
Eaton, J., Kortum, S. S., & Kramarz, F. (2022). Firm-to-Firm Trade: Imports, Exports, and the Labor Market (NBER Working Paper Series No. 29685). National Bureau of Economic Research.
Miyauchi, Y. (2023). Matching and Agglomeration: Theory and Evidence from Japanese Firm-to-Firm Trade [Working Paper]. Boston University.
Oberfield, E. (2018). A Theory of Input-Output Architecture. Econometrica, 86, 559–589.
OpenStreetMap contributors. (2023). Costa Rica Map Dump. https://www.openstreetmap.org